hi friends!

i'm pretty tired of engagement bait and all the political nonsense on my x/twitter feed.

i was curious if i could use an llm to filter out these type of content, so i prototyped a quick chrome extension.

it uses LLama 3.3 to analyze the tweet through https://groq.com/ (because they are super-super fast).

the extension is available in the chrome store, also there is a link to the repo.

- you can tweak the system prompt for the filtering - but you need your own API key from Groq (you can get one for free)

  • nomilk 6 days ago |
    I'm personally a huge Elon fan, but I know many who are not and have heard from multiple sources that they can't get him out of their feeds no matter how hard they try. If you tweak your current idea to simply remove all Elon-related content from the feed, that would be funny but may also be strangely profitable (people would probably buy it out of spite).
    • plaguuuuuu 6 days ago |
      [flagged]
      • nomilk 6 days ago |
        Leaving X is one way. But there is an intersection in the venn diagram between those who find X valuable and those who dislike Elon's content.
      • ImJamal 6 days ago |
        You won't get messages from Elon, but you will sure get a lot of messages about Elon since he lives rent free in so many people's mind.
        • palmfacehn 6 days ago |
          At the time of posting, I count 8 matches for "Elon" on this page and 23 comments. All of but one of the mentions has a negative sentiment.
          • croisillon 6 days ago |
            how dare they talk (badly!) about him in a thread about twitter!
            • palmfacehn 6 days ago |
              The above poster didn't elaborate when he mentioned, "Living rent free". If we are going to have a discourse on the topic, I think it deserves more depth.

              He's a complicated figure. I'm not a fan of all of his ideas, nor do I consider him some kind of right-wing free speech savior. He's like anyone else in that there's some hit and miss. I do find it disturbing how quickly the predominantly left-leaning tech sphere turned on him. This coincided with politically driven media talking points.

              A few short years ago he was celebrated for his association with green energy. He was even celebrated by the globalists at the WEF. For this and other reasons, I suspect his promotion to be something of a false alternative. Idolizing him seems a bit naive. The demonizations I regularly read on this site are often poorly reasoned or Godwinian hyperbole.

              • croisillon 6 days ago |
                is... is "globalist" a code word?
                • palmfacehn 6 days ago |
                  https://www.wita.org/atp-research/free-trade-ideology/

                  In short, it is the philosophy espoused by the WEF, The Club of Rome and some UN institutions.

                  For all of Twitter's faults, I never interact with someone making these toxic insinuations on that platform.

              • felixgallo 5 days ago |
                can you provide an example of the 'politically driven media talking points' that unfairly stained poor Elon's otherwise sterling reputation with the green, ah, globalists?
          • ImJamal 6 days ago |
            To be clear, I think people who love Elon also have a problem. Neither side can stop thinking and talking about him. I would love to go one day without somebody talking about him on HN regardless of the sentiment.
        • delecti 5 days ago |
          He's the owner of the site, in the inner circle of the incoming administration, and also the richest person in the world as global society is increasingly an oligarchy. That's hardly "living rent free in [one's] mind".
          • ImJamal 5 days ago |
            People were nonstop talking about him prior to being involved with Trump and Twitter and it has only gotten worse.
  • patchtopic 6 days ago |
    [flagged]
    • blueflow 6 days ago |
      The platform was not the problem, it was the people. People migrated to bluesky and fedi and now create these contents there.
      • anthk 6 days ago |
        Elon was one of the main enshittifiers. I even used to read Twitter without JS.
      • input_sh 6 days ago |
        If the platform wasn't the problem, nobody would've moved.

        Closing down the API, charging for the checkmark, not displaying tweets to those not logged in, artificially boosting tweets by the owner itself etc etc.

    • xmodem 6 days ago |
      Switching to mastodon has been such a breath of fresh air. No algo-driven feed manipulating me to constantly check in means that I have a much healthier relationship with it. If I'm, say, on the metro I can use it to kill time for 5 minutes... and then put it down, and I don't feel compelled to pick it up.
      • bArray 6 days ago |
        Be careful what air you breath in: "Twitter rival Mastodon rife with child-abuse material" [1] [2]. A lot of this was previously limited by X until a mass exodus to platforms such as Mastodon and BlueSky.

        [1] https://www.washingtonpost.com/politics/2023/07/24/twitter-r...

        [2] https://archive.is/SNYcK#selection-423.0-423.53

        • GaryNumanVevo 6 days ago |
          Actually this is a big reason why a lot of instance owners block replication of any posts from Japanese instances
      • Kiro 5 days ago |
        Every time I see a post from Mastodon it's always some hate and rage driven rant. Like "I can finally say what I want so f* this and that". Anecdotal of course but it doesn't really come off as less toxic.
        • xmodem 5 days ago |
          What you can do, is curate who you follow / block so that you mostly don't see that. TBF you could do the same thing on X/Twitter to an extent but the platform has an incentive to circumvent your preferences.
  • yellow_lead 6 days ago |
    Its impressive how fast this is. I don't use twitter anymore (for the same reason as your use case), but nice job!
    • wkat4242 6 days ago |
      It's groq, they have amazing custom built hardware.
    • danielpetho 6 days ago |
      yes groq is really amazing.

      there is another player i've found when making this -> https://inference.cerebras.ai/

      never tried their api tho

  • eddieh 6 days ago |
    It was much easier and healthier for me to just leave Twitter (never gonna call it the other name). In fact, I’ve left nearly all of social media. It would be nice to get away from all algorithmic content feeds. Maybe someday.
    • nomilk 6 days ago |
      Social media apps have downsides, but I find the algorithmic short videos on facebook ("reels") quite incredible. The algo seems to suggest mostly science; niches I would never have known about, let alone had an active interest in, had the algo not suggested them.

      I suspect experiences using social media apps differ wildly from person to person.

      • enaaem 6 days ago |
        X is by far the least fun social media app. I tried it and I never got the appeal. Just stick to insta for lighthearted content.
        • nomilk 6 days ago |
          Fun and useful are different things though. I use X for 'news', debate, and updates on anything global. YouTube for specific tech tutorials and 'junk' viewing (e.g. something light-hearted while eating). And insta/fb reels for filling in idle gaps (e.g. while waiting in a queue). I gave up legacy media about 3-4 years ago and although Twitter is trash, would take it over mainstream media any day of the week (same with BlueSky and Mastodon, but I simply haven't needed those since they're approximately equivalent to X).
          • pavlov 6 days ago |
            Didn’t we use to laugh at boomers who got their news from their Facebook feed?

            X is the same thing, but with the lies amped up 10x and fractured into discontextualized atoms of outrage to keep you scrolling.

            • nomilk 6 days ago |
              Yup. The variance of truthiness on X is probably the highest of all platforms. Meaning you get the most accurate news if you can filter/discern well enough, at the risk of being brainwashed if you can’t. Since alternatives also brainwash, X is more or less pareto optimal
              • pavlov 6 days ago |
                There is a heavy Dunning-Kruger effect where people think they are very discerning and sophisticated in how they filter, but gradually descend into depths of misinformation without realizing.

                Grandparents sharing nonsense on Facebook also think themselves highly informed — after all they’ve spent the past ten years in retirement “doing their own research”.

              • mrguyorama 5 days ago |
                >The variance of truthiness on X is probably the highest of all platforms. Meaning you get the most accurate news if you can filter/discern well enough

                Absolute nonsense. Believing X is a simple distribution or sample of reality is bonkers.

          • soco 6 days ago |
            I guess people like me have zero interest in global news, debate and updates on anything global. Or rather, not zero interest, but zero time for it after filling the day with other (more important to me) things.
        • dustingetz 6 days ago |
          i followed 1000 scientists on X last year. Quantum mechanics, particle physics, general relativity, astrophysics and cosmology, molecular bio, chemistry, math, microbiology, environment, linguistics, psychology, middle east, islam, French. Oh and work stuff: programming languages, compilers, AI, venture capitalists of all stage. Where else can you find this? The main issue with X is it is a mirror. It puts us in a room with people like ourselves. If you like to argue, it will send you a lot of bait. I just stopped doing that. Muted a few keywords (nazi, greta, trump, elon, gun). And read a lot of paper summaries about black holes! (Oh and unfollowed most of the VCs, that was terrible)
      • vintermann 5 days ago |
        I had an amazing experience with Facebook shorts! What happened was I...

        I had an amazing experience with Facebook shorts! What happened was I...

        I had an amazing... (hits the close button)

        • nomilk 5 days ago |
          That misses the point.

          Consider an analogy with a knife. If someone quits using knives simply because they had previously cut themselves, and someone says: “wait, there’s at least one useful application for these things”, and demonstrates the useful application, that could be enough for that person to discover how useful knives can be when used right.

          Exact same goes for any tool that can cause both harm and benefit, like social media apps.

          • vintermann 4 days ago |
            What do you think my point was?

            I was just making fun of desperate engagement baiting Facebook short videos that of people just about to get kicked by a camel or whatever, looping right before whatever they're baiting actually happens.

      • BeFlatXIII 5 days ago |
        > I find the algorithmic short videos on facebook ("reels") quite incredible

        Mine discovered that I go wild for horse videos. The fewer humans on screen, the better.

    • bjornsing 6 days ago |
      It pains me but I’m leaning in this direction too. I found the idea of a digital town square with a commitment to maximum truth appealing, but seeing what Twitter has become it’s hard to not feel this was naive of me.

      To me the big blow was seeing the response to what’s happening in Gaza and what narratives people and algorithms in combination end up promoting. The thoughtful, balanced, humanistic view gets approximately zero traction, while completely untruthful propaganda (on both sides) has enormous reach.

      Maybe it’s an insurmountable problem. Human defense mechanisms in combination with algorithms will always push people to tribalism and cheering for atrocities. I hope not, but seems like it.

      I guess you could take the OPs approach of filtering out all the propaganda and keep contributing. But then you are effectively working for a propaganda machine free of charge, helping create value that will draw others in to be subjected to the propaganda.

      • soco 6 days ago |
        "completely untruthful propaganda" honest question: how are you able to discern that?
        • bjornsing 6 days ago |
          It’s often not too hard to determine if you put in the effort. For example you have big accounts like John Spencer spreading claims that Gaza’s population growth is completely unaffected by the “war”, with thousands of likes and hundreds of thousands of views [1].

          If you need an example on the other side you can take this popular post that claims the German foreign minister has said Israel has a right to target civilians [2], when in fact she says Israel has a right to target civilian infrastructure, if it is being used for military purposes.

          Community notes do not appear because the algorithms require people who typically disagree to agree, which I doubt will ever happen in a military conflict.

          1. https://x.com/spencerguard/status/1864551769247723645

          2. https://x.com/occultni/status/1875808201477943362

    • alias_neo 6 days ago |
      I only ever used Twitter for news because it was were all of the big names were;, but I stopped using it entirely when it took a massive dive after the recent sale. I checked back a couple of days ago, and top of my feed was some BS take on a subject (I forget what), and _every_ _single_ response was from a blue-tick agreeing with it, I immediately knew I can close it and not come back unless I want to click through to a specific tweet for whatever reason.
  • jamesrcole 6 days ago |
    I think this is a glimpse of the future. I can imagine ever-greater amounts of information overload, and AI used everywhere to filter it to manageable levels.
    • freehorse 6 days ago |
      This sounds like an extreme waste of resources in all sides.
      • bryant 6 days ago |
        Extreme is right; I can only imagine the waste of power. It's probably exponentially worse than the waste of gas by postal services for delivering junk mail.

        Except I'm not sure there's anything that can be done about it in the US because of first amendment rights.

        • freehorse 6 days ago |
          People (or "people" aka bots) do not write bs because they can. They write because they are incetivised to do so.
      • input_sh 6 days ago |
        LLMs in a nutshell.

        Also see: people using them to write artificial garbage and then readers using them to summarise the said garbage.

    • wkat4242 6 days ago |
      And on the other side AI being used to create more of it and to find ways to bypass the filters.
  • anthk 6 days ago |
    Just leave out Xitter. Read blogs over RSS and gopher phlogs (SDF, bongusta...)

    Also, Usenet and Dove/Fidonet (you can get them over NNTP news://cvs.synchro.net:119) have more polite discussions on politics.

  • palmfacehn 6 days ago |
    90% of my feed is just AI generated art. The mute button seemed to work on the baiters.
    • tomalbrc 5 days ago |
      *AI generated pictures
  • dewey 6 days ago |
    If you have to fight this much to use a platform...why bother?
    • danielpetho 6 days ago |
      the network effect is still strong.

      it feels like a huge chunk of (recent/real-time) tech/design discussion happens there. but could be just me. i'm still warming up on bluesky/threads.

      • pwdisswordfishz 5 days ago |
        Why do you insist on perpetuating it?
  • wkat4242 6 days ago |
    Um yeah good idea in principle but for me if I don't want to see the bad content it's better to just not look at it. The hate is everywhere, even on positive posts in the comments.

    But good idea yes.

    • danielpetho 6 days ago |
      i agree. but sometimes my lizard brain just can't resist the temptation and i will stop to read the tweet/start a video. and the algorithm already could think that i'm interested in this type of content
  • gigatexal 6 days ago |
    Would instantly use it if I could also have it filter Elon’s tweets and his fans and followers.
    • danielpetho 6 days ago |
      you might be able to craft a system prompt eg.: filter out anything that mentions or even slightly related to elon musk, his companies, etc.

      obviously it's never going to be perfect tho

      but your best bet is still blocking/muting people then

  • DantesKite 6 days ago |
    Such a great demonstration of the concept.

    I think it’s inevitable that we’ll start to see more sophisticated ways of organizing our social media feeds.

    I don’t think it has to be this binary decision where we either abandon social media altogether or expose ourselves to the most emotionally draining content possible. There’s likely many different unexplored metas as it were.

    I often joke that we should have a marketplace of algorithms we can subscribe to, where the sentiment “slider bar” can go from Hello Kitty Island Adventure positivity to 4Chan LiveLeak nihilism, if you so choose.

    • danielpetho 6 days ago |
      yeah some kind of real-time tweaking would be also nice. sometimes i'm in the mood for topics i'm usually not following, but i would go back to my "default" space after a while.

      the problem imo is that there is absolutely zero control on most platforms.

      on x there is a button called "i'm not interested in this tweet" but obviously it doesn't do anything meaningful

    • MattDaEskimo 5 days ago |
      This is what the algorithm & tools of the platform is supposed to do for you.

      It's also what makes you valuable as a free consumer.

      If you want lumber, don't buy a house and strip it

  • beAbU 6 days ago |
    I've found this [1] to also work really well to clean up the feed.

    1: https://help.x.com/en/managing-your-account/how-to-deactivat...

    • buran77 5 days ago |
      I went that route for my account which I wasn't using anyway but only because I realized it's not "my account" and my control over it is way too limited.

      I logged in after many years of inactivity only to realize that despite only having a few friends, my feed was exclusively Musk and Tesla stuff. No amount of "don't show me this" helped. On one hand I found it exceptionally pathetic that any person would need to stroke their own ego to this degree at the expense of everyone else. But worse, it's clear that Twitter/X/Musk control my feed more than I do so the only way to win the game is to not play it.

    • thih9 5 days ago |
      The link leads to a Twitter/X support article for “How to deactivate your account”.
      • sphars 5 days ago |
        That's the joke...
        • thih9 5 days ago |
          I know; my comment saves time (at least on devices where the link is truncated).
  • insin 6 days ago |
    > i'm pretty tired of engagement bait and all the political nonsense on my x/twitter feed

    Getting rid of "For you" entirely and removing Retweets from the chronological timeline (default configuration for [1]) gets you 95% of the way there - you'll only get engagement bait and political nonsense if the people you choose to follow are the ones actively doing it, in which case…

    [1] https://jbscript.dev/control-panel-for-twitter

    • danielpetho 6 days ago |
      oh wow, this looks promising. thx!
    • ksec 5 days ago |
      Yes. I dont even use "For you". I dont even "follow" most people, just put them into a list and visit that list only. In hind sight I have been using X / Twitter like RSS for a very long time. I guess that is because coming from RSS and I dont know how Twitter work at first and I first thought why do I have all these random people on my feeds.

      I have lists for Software, Hardware, Politics, VCs, Economy, Art, and some other wired stuff.

      You can ignore 99% of Spam, bait etc and only focus on the signal.

  • yomismoaqui 6 days ago |
    Already implemented something similar but using Gemini Flash 1.5

    https://twitter.cercaapp.com

    • danielpetho 6 days ago |
      neat!

      yes it's the same idea, tho for this one you need your own API key, which can be a bummer for some.

  • strajk 6 days ago |
    Nice! I'm glad I was not the only one frustrated by these and I actually made sth similar just this weekend, also using Groq cause it's super fast as u said. (but haven't published it cause I'm always frustrated about that part of sideprojects :D) https://x.com/Straaajk/status/1875492276715315405
    • danielpetho 6 days ago |
      super cool! i like yours better tbh as it looks much more granular in terms of categorizing

      re publishing—just do it! this project is also a rough mess, only a quick experiment, so i doubt anybody will care if it fails sometimes / not perfect, code is a mess etc.

  • haunter 6 days ago |
    I don't have to because I just use the "following" tab? I only see people I follow. Really don't get the point of this service
    • danielpetho 6 days ago |
      i like the for you page since i can discover fresh/new work from others. otherwise how would i get to the point of following someone? (excluding when you find the person's account on an external site)

      but i feel you, the following tab is definitely a healthier/more controllable space

      • geerlingguy 5 days ago |
        For years before the algorithm-generated pages were defaulted, you would find new accounts organically—through retweets and conversations.

        I still do it that way, better to do a (very) slow roll and build up a network slowly than to follow the accounts Twitter's algorithm thinks I should follow (IMHO).

      • andrewla 5 days ago |
        Your "following" feed contains content liked or retweeted/replied to by people you follow. At least, it used to -- since likes went private I'm not sure that you see this any more.
  • vanrohan 5 days ago |
    I'm also on a mission to clean the "For You" feed. I tried using local LLMs to do this real-time classification, but it's currently just too slow.

    Another way to reduce the spam, is to remove bot followers from your account, I believe these bots use "likes" (which are hidden) to boost content for the botnet owners. Most of these bots are easy to spot, 0 posts, and very bad follower/following ratios. I built this Chrome Extension for bot removal automation: https://chromewebstore.google.com/detail/x-bot-remover/aohkh...

    It would be way better if this was just done by X themselves, but why not try.

    • danielpetho 5 days ago |
      yeah, something local (and free) would be much better. i'm not even sure how expensive could this be if it would run all the time...

      re bot removal: thanks for it, will give a it a try!

    • russdill 5 days ago |
      It'd be nice if x would let you do what bluesky does, let you write your own algorithms and create your own feeds.
  • pluc 5 days ago |
    If you don't want engagement bait or politics on Twitter.. why are you on Twitter? RSS is better to keep up to date on news, and if none of your friends are on Mastodon, get new friends.
    • bearjaws 5 days ago |
      I think Twitter has two core users

      1. People who want to debate / react to things / politics / just get angry apparently.

      2. People who are like a town crier from times past, they want to be the first to know some event has happened and then go to their group chats & other social media to post about it.

      • danielpetho 5 days ago |
        i disagree, or at least i don't consider myself to either of these groups.

        i think there is still a large enough group which is there to share work / projects

      • MattDaEskimo 5 days ago |
        3. Those who just don't care about the sociological trends and impact of using, and therefore promoting Twitter
    • andrewla 5 days ago |
      Sorry, ignorant question -- how does one us RSS to keep up to date on news?

      With Twitter, I can see content from people I follow and from content referenced or highlighted by people I follow. People in this case can also include news organizations or other professional journalistic content.

      In addition there is the algorithmically generated garbage feed stuff that is mostly useful for seeing (potentially) interesting content that I missed on the main feed, but also contains a ton of content from people I did not follow and that people I do follow did not highlight or reference. Presumably this project attempts to reduce the garbage-ey ness of this feed, which I would appreciate.

      • pluc 5 days ago |
        For the next month, every time you find something interesting on Twitter, go to the source. Look for their RSS feed - if they don't have one, ask them to work on implementing one both via Twitter and via their contact form.

        I don't know if it's still the case, but there used to be RSS feeds for Twitter accounts too.

        • andrewla 5 days ago |
          And what then do I do with the RSS links? The last time I dealt with RSS at all was in the Google Reader days. Are there cross-platform solutions that maintain any context? Is there any way to reply or give feedback through RSS, or participate in conversations?
          • serenadeineb 5 days ago |
            check out feedly.com its a nice RSS aggrogater website, with a competant ios app as well. Ive used it forever, and like the parent said, when I find myself on a site that provided a story I found interesting, I tried to add the host website ( like https://example.com ) as a rss source, and feedly checks and offers any rss feeds it finds.
          • pluc 5 days ago |
            I use Feedbro for Firefox.

            You can comment wherever people listen to you.

    • akopkesheshyan 5 days ago |
      Completely agree. No need to over engineer problem. Just leave the network
  • Havoc 5 days ago |
    Prompt used seems to be here:

    https://github.com/danielpetho/unbaited/blob/main/extension/...

    Seems like a reasonable idea. I have doubts about an LLM being able to reliably detect when a tweet was made to *deliberately* trigger emotions. That’s setting the bar quite high on llm working out intent

    • danielpetho 5 days ago |
      100% agree unfortunately, haha

      the default system prompt is quite weak, and could be much more specific / better.

      also, even as a "human" can't decide sometimes what is a bait...

  • bentt 5 days ago |
    Too many replies are missing the point. This isn't about whether there are other ways to do this, or if X is good/bad. This is about using a local AI model to curate content we want to avoid/see. This is a big area of interest and I'm happy to see something on HN about it!
    • ikanreed 5 days ago |
      But, of course, it would be better if we could generalize and let users describe in conversational language the specific things they don't want to see.
    • itslennysfault 5 days ago |
      It's not a local AI model. It's using a groq.com API, but your point about people missing the point is still a good point.
    • danielpetho 5 days ago |
      yes that's the main idea.

      i mean even llms wouldn't be necessary (this is just a possible direction), but just give me more control on my own feed

  • derivagral 5 days ago |
    Cool idea! I had something similar for using LLMs as an intermediary not for spam but for "boring" local news matching to your profile. Court stuff, town meetings on youtube, that kind of thing.
    • danielpetho 5 days ago |
      sounds interesting! tho that would be rather summarizing content i assume?
  • andy_ppp 5 days ago |
    I want this for YouTube videos as well that can summarise the content and just take me to the things I would find most interesting in the video, almost like auto chapters that skip all the “and later we’ll be discussing really interesting thing X if you keep listening” filler. It would be great to filter all the noise out and create highlights package automatically that was 10% of the video length that can be watched separately.
  • andrewla 5 days ago |
    Tangential, but I think I speak for many when I say that we no longer need to specify "X" as a name for twitter -- just call it twitter and tweets and ignore the abysmal rebrand until the day that it gets reverted.
  • j45 5 days ago |
    Filtering out is a nice feature.

    Highlighting each piece as such and teaching as what about it is bait, etc would make people stronger and more aware.

  • pjmlp 5 days ago |
    I have been ramping down my use until I collect the info I need to follow specific people elsewhere, then I am out as well.
  • ryan_lane 5 days ago |
    Why do you use twitter if you want to log into a completely blank feed every time you use it?